Skripsi

MODEL MULTICLASS CLASSIFICATION UNTUK PENYAKIT BERDASARKAN CITRA CHEST X-RAY PARU-PARU DENGAN ENSEMBLE LEARNING.

---

Penilaian

---

Chest X-ray (CXR) images can diagnose lung diseases. However, diagnosis requires time and high accuracy, so an automated system is needed. In the process, the CXR image is first enhanced for image quality using Morphology Opening and Median filter, followed by data augmentation using rotation and flipping. CXR image segmentation using U-Net Batch Normalization is done by separating the lung object from the background. The results of the segmentation process are carried out implementation of the ensemble learning method on the performance of ResNet, EfficientNet, and Inception-v3 architectures. The results successfully improved the quality of CXR images using the morphology opening and Median filter, with an average PSNR value of 39.307, an MSE of 22.469, and an SSIM of 0.952. The U-Net Batch Normalization segmentation model achieved an accuracy of over 93% and a loss value close to 1%, indicating an excellent ability to detect lungs in CXR images. The application of Ensemble Learning from ResNet, EfficientNet, and Inception-v3 (ELREI) in the classification stage resulted in significant improvement compared to the single classification method. The increase in accuracy value is 11% (ResNet), 3% (EfficientNet), and 1% (Inception-v3). The increase in precision value is 10.5% (ResNet), 1% (EfficientNet), and 3% (Inception-v3). The increase in recall value is 10.75% (ResNet), 1% (EfficientNet), and 3.25% (Inception-v3). The increase in F1-Score value is 10.25% (ResNet), 3.25% (EfficientNet), and 1% (Inception-v3). The average accuracy, precision, recall, and F1-Score generated by the ELREI method are 99%, 98.75%, 98.75%, and 99%. Overall, the ELREI method proved to be robust and excellent for classifying lung diseases based on CXR images by categorizing them into four classes: COVID-19, normal, lung opacity, and pneumonia.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2407001561	T129847	T1298472023	Faculty of Economics (Referens)	Available

Detail Information

Series Title

-

Call Number

T1298472023

Publisher

Palembang : Prodi Doktor Ilmu Teknik, Fakultas Teknik., 2023

Collation

xiv, 91 hlm.; Ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

621.390 7

Content Type

Text

Media Type

unmediated

Carrier Type

-

Edition

-

Subject(s)

Ilmu Teknik, Teknik Komputer

Specific Detail Info

-

Statement of Responsibility

MURZ

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment